THE UNIVERITY OF TEXA AT AN ANTONIO, COLLEGE OF BUINE Working Paper ERIE Dae May 15, 013 WP # 0046FIN-0-013 Commodiy Financializaion and Herd Behavior in Commodiy Fuures Markes Rıza Demirer Deparmen of Economics & Finance ouhern Illinois Universiy Edwardsville Edwardsville, IL 606-110 Hsiang-Tai Lee Deparmen of Banking and Finance, 1 Universiy Rd., Puli, Nanou Hsien, Naional Chi Nan Universiy, Taiwan 54561 Donald Lien Deparmen of Finance Universiy of Texas - an Anonio One UTA Circle an Anonio, TX 7849 Copyrigh 013, by he auhor(s). Please do no quoe, cie, or reproduce wihou permission from he auhor(s). ONE UTA CIRCLE AN ANTONIO, TEXA 7849-0631 10 458-4317 BUINE.UTA.EDU
Commodiy Financializaion and Herd Behavior in Commodiy Fuures Markes Rıza Demirer Deparmen of Economics & Finance ouhern Illinois Universiy Edwardsville Edwardsville, IL 606-110 Hsiang-Tai Lee Deparmen of Banking and Finance, 1 Universiy Rd., Puli, Nanou Hsien, Naional Chi Nan Universiy, Taiwan 54561 Donald Lien Deparmen of Finance Universiy of Texas - an Anonio One UTA Circle an Anonio, TX 7849 May 013 E-mail: rdemire@siue.edu; Tel: 618650939; Fax: 1618650-3047. Please do no quoe wihou permission. 1
Commodiy Financializaion and Herd Behavior in Commodiy Fuures Markes Absrac This paper conribues o he debae on commodiy financializaion by exending ess of herd behavior o he commodiy fuures markes. Uilizing a regime-swiching model, we es he presence of herd behavior in a number of commodiy secors including energy, meals, grains and livesock during he low and high marke volailiy saes. We find significan evidence of herd behavior in grains only during he high volailiy sae. We also find ha large price movemens in he energy and meals secors significanly conribue o herd behavior in he marke for grains. Finally, we find no significan effec of he sock marke on herd behavior in he commodiy fuures marke. Our findings in general do no suppor he much debaed commodiy financializaion hypohesis. JEL Classificaion Code: G14, G15 Keywords: Herd behavior, Commodiy financializaion, Reurn dispersion, Markov swiching
1. Inroducion peculaion in commodiy markes has been he source of heaed discussions among policy makers as well as in he media. Paricularly, he 008 bubble in he prices of a wide range of commodiies has focused policy makers aenion o he role of financial invesors aciviies in commodiy markes. Echoing George oros saemens in a esimony before he U.. enae Commerce Commiee Oversigh Hearing on FTC Advanced Rulemaking on Oil Marke Manipulaion 1, Gilber (009) suggess ha a new class of invesors ha has emerged in financial markes regard commodiies as an asse class, comparable o socks, bonds, real esae, and emerging marke asses, and ake posiions on commodiies as a group in order o capure profis ha are no possible o obain from radiional asses. Amazingly, a he peak of he price bubble in 008, commodiy fund invesors, including hedge funds like oros Fund Managemen, conrolled a record 4.51 billion bushels of corn, whea and soybeans hrough he fuures markes of Chicago Board of Trade, equal o half he amoun held in U.. silos on March 1, 008. In a esimony before he U.. enae Commiee of Homeland ecuriy and Governmen Affairs, Michael W. Masers, a porfolio manager and parner a he Masers Capial Managemen, LLC saed: 3 You have asked he quesion Are insiuional invesors conribuing o food and energy price inflaion? And my unequivocal answer is YE. In his esimony before he U.. enae Commerce Commiee, George oros also saed ha commodiy invesmen, as a new venue for insiuional invesors, had become he elephan in he room and as a resul, invesmen in hese asses migh exaggerae price rises. To his end, a number of sudies on financial markes have suggesed ha herd formaion among large 1 oros, G. (008), Tesimony before he U.. enae Commerce Commiee Oversigh Hearing on FTC Advanced Rulemaking on Oil Marke Manipulaion, Washingon D.C., 4 June 008. Wilson, J. (008), Wall ree Grain Hoarding Brings Farmers, Consumers Near Ruin, Bloomberg, (April 8, 008) 3 Masers, M.W. (008), Tesimony before he U.. enae Commiee of Homeland ecuriy and Governmen Affairs, Washingon, DC, 0 May 008. 3
insiuional invesors may desabilize marke prices and creae excess volailiy (Dennis and rickland, 00; Luo, 003; Gabaix e al., 006). Therefore, one may argue ha herd behavior in he commodiy marke, possibly driven by financial invesors moving funds in and ou of commodiies, is a conribuing facor behind he booms and buss observed in a wide range of commodiies. On he oher hand, sudies including Krugman (008), Hamilon (009), and Kilian (009) rejec he so-called commodiy financializaion hypohesis and sugges ha commodiy price cycles are mainly driven by supply and demand balances in global markes, largely due o growh rends in emerging economies. Adding suppor o his view, Buyuksahin and Harris (011) examine he rading posiions of various ypes of raders in he crude oil marke and find lile evidence ha financial invesors posiion changes cause price changes in he oil marke. Given he conflicing views in boh direcions, invesor behavior in he commodiy marke and how i relaes o he excessive price movemens is ye o be explored. The main goal of his paper is o conribue o he discussion on he financializaion of commodiies from a differen angle by exending ess of herd behavior o commodiy fuures markes. Uilizing a mehodology applied o a number of financial markes, we examine price daa from energy, grains, livesock, and meals fuures and es wheher herd behavior is presen during he low and high marke volailiy saes. Our findings sugges he presence of herd behavior in he marke for grains only wih no evidence of herding in oher commodiy secors. Herd behavior in grains is observed during he high marke volailiy sae only. Furhermore, he resuls do no sugges a significan effec of sock marke movemens on herding in commodiy markes, hus providing evidence agains he commodiy financializaion hypohesis. On he oher hand, a significan cross-marke herding effec on grains is observed from he energy and meals markes, suggesing ha large price movemens in energy and meals end o conribue o herding among invesors in grains fuures. Our findings are robus during he pos-004 period 4
when he commodiy marke experienced a large influx of financial invesors driving a dramaic rise in open ineres and rading volume in commodiies (Figure 1), furher supporing evidence agains he commodiy financializaion hypohesis. An ouline of he remainder of he paper is as follows. ecion summarizes he lieraure on herd behavior. ecion 3 provides he deails of he esing mehodology and daa descripion. ecion 4 presens empirical resuls. Finally, ecion 5 concludes he paper.. Previous udies A number of sudies in he lieraure have examined herd behavior in differen markes and using alernaive mehodologies. Chrisie and Huang (1995) describe herd behavior as a endency for individuals o suppress heir own beliefs and base heir invesmen decisions solely on he collecive acions of he marke, even when hey disagree wih is predicions. Bikhchandani and harma (001) define herding as an obvious inen by invesors o copy he behavior of oher invesors and buy and sell an asse as a group. udies including hleifer and ummers (1990), Avery and Zemsky, (1998), and Chari and Kehoe (004) propose an informaion based heory for herding where individual invesors follow he signals from he rades of more informed agens wih beer access o informaion compared o individual invesors. Devenow and Welch (1996) sugges ha managers in an imperfecly informed marke may prefer eiher o hide in he herd no o be evaluable, or o ride he herd in order o prove qualiy. Oher sudies including charfsein and ein (1990), Rajan (1994), Graham (1999), and wank and Visser (008) sugges ha fund managers imiae ohers as a resul of he incenives provided by he compensaion scheme or in order o mainain heir repuaion. Neverheless, whaever he raionale behind such behavior may be, sudies including Dennis and rickland (00), Luo (003), and Gabaix e al. (006) sugges ha herd behavior may lead o excess volailiy by leading asse prices deviae from fundamenal values. 5
The lieraure offers an exensive lis of sudies on herd behavior applied o a number of differen markes. A commonly used esing mehodology ha is based on asse reurn dispersions is uilized in Chrisie and Huang (1995) on U.. equiies, Chang e al. (000) on inernaional equiies, Gleason e al. (003) on commodiy fuures raded on European exchanges, Gleason a al. (004) on exchange raded funds, Demirer and Kuan (006) and Tan e al. (008) on Chinese socks, Demirer e al. (010) on Taiwanese socks, Chiang and Zheng (010) on global sock markes, and more recenly Philippas e al. (013) on REITs and Balcilar e al. (013) on Gulf Arab sock markes. However, hese ess have no ye been exended o U.. commodiy fuures. Regarding sudies on commodiy markes, saring wih Pindyck and Roemberg (1990), several sudies have suggesed ha herding among raders may lead o excess comovemens among commodiy prices. Wiener (006) examines speculaive behavior in he inernaional oil marke in he mid-1990s and finds ha some subgroups of invesors end o ac in parallel in heir rades. imilarly, Gilber (009) disinguishes beween speculaors and commodiy funds and finds some evidence of shor-run explosive behavior in non-ferrous meals markes due o speculaive aciviies. On he oher hand, Chunrong e al. (006) rejec speculaion and herding as he source of commodiy price comovemens, providing evidence agains herding in commodiy markes. Adrangi and Charah (008) acknowledge some degree of relaion among he posiions of commodiy raders, however heir resuls show he relaedness falls shor of herding. imilarly, Boyd e al. (009) examine rading daa and find ha herding among hedge funds does no desabilize he crude oil marke. More recenly, een and Gjolberg (013) examine he correlaion paerns and principal componens describing commodiy reurns in order o make inferences on herd behavior and find no significan suppor. In shor, he lieraure provides conflicing evidence on herd behavior in commodiy markes. Ineresingly, he reurn dispersion based mehodology ha is used exensively in he lieraure o es he presence of herd behavior 6
has no ye been exended o U.. commodiy fuures markes. The only excepion is Gleason e al. (003) who examine commodiy fuures raded on European exchanges and find ha raders in European fuures markes do no have herding endencies. To he bes of our knowledge, his sudy is he firs o exend reurn dispersion based herding ess o U.. commodiy fuures. 3. Daa and Mehodology 3.1 Daa The daase consiss of weny fuures conracs: five energy (crude oil, heaing oil, naural gas, gasoline and ehanol), four livesock (feeder cale, live cale, lean hogs, and pork bellies), six grains and oil seeds (whea, corn, soybeans, oas, rapeseed and rough rice), and five meals (gold, silver, plainum, palladium and copper). Daily nearby fuures prices covering he period beween Jan. 17, 1995 and Nov. 30, 01 are obained from Commodiy ysems Inc. The reurns for he nearby monh fuures conracs are uilized in he ess. Nearby fuures prices are consruced wih conrac rollover occurring abou one week before he mauriy in mos cases. The rading volume is used as a crierion in deciding he acual rollover dae. 3. Mehodology We follow a commonly uilized mehodology o deec herding behavior in financial markes. Originally suggesed by Chang e al. (000), he esing mehodology focuses on he relaion beween he dispersion of asse reurns wihin a porfolio of asses wih similar characerisics and marke movemens. Dispersion of reurns wihin a porfolio is measured by he cross-secional absolue deviaion of reurns (CAD) expressed as 1 N CAD N i1 R i R, m. (1) where N is he number of asses in he porfolio, R i, is he reurn on asse i for day and R m, is he daily reurn on a measure of he overall secor. Bikhchandani and harma (001) sugges 7
ha herding behavior would be more likely o occur a he level of invesmens in similar asses where invesors face similar decision problems and can observe he rades of ohers in he group. For his purpose, we organize each fuures conrac ino four commodiy secors, i.e. energy livesock, grains and meals. Each commodiy secor is represened by he corresponding &P GCI index. The reurn dispersion measure in Equaion 1 can be regarded as a proxy o individual asse reurn dispersion around he marke reurn. From an efficien marke perspecive, one would expec reurn dispersion o increase wih he absolue value of marke reurn since each asse in he porfolio differs in is sensiiviy o marke shocks. However, Chang e al. (000) argue ha he presence of herding behavior would lead asse reurns no o deviae far from he overall marke reurn. In oher words, he correlaed acions of raders as hey suppress heir own beliefs and make invesmen decisions based solely on he collecive acions of he marke, would lead asse reurns o display greaer direcional similariy, hus leading o lower reurn dispersion wihin he commodiy porfolio. ince such an invesmen behavior would be more likely o occur during periods of marke sress characerized by large price movemens, Chang e al. (000) propose a esing mehodology based on a general quadraic relaionship beween reurn dispersion and marke reurn in order o deec herd behavior. In his sudy, we esimae a generalized version of he model by Chang e al. (000) which accouns for he GARCH effecs in he ime series and esimae for commodiy secor k CAD e h k, 0 1 Rk., ~ 0,1 1 N 0 1 1 1 R k. e h h e () where CAD k, is he cross-secional absolue deviaion of fuures conrac reurns in commodiy secor k and R k, is he reurn on commodiy secor k on day. In his specificaion, h sands for he condiional variance assumed o follow a sandard GARCH(1,1) process. According o he 8
esing mehodology, herding would be evidenced by a lower or less han proporional increase in he cross-secional absolue deviaion (CAD) during periods of large price movemens. As a resul, observing a negaive and saisically significan α would be consisen wih he presence of herd behavior. A significan weakness of he model in Equaion () is ha i is saic in naure, i.e. he parameers are assumed o be consan over ime, ignoring possible srucural breaks. Therefore, he model fails o differeniae marke saes during which herding behavior may or may no be presen. For his purpose, we exend he saic model in Equaion () o a regime-swiching framework and differeniae beween low and high marke volailiy saes. If invesors are more likely o herd during periods of high marke volailiy, hen a regime-based model should be able o idenify herding and non-herding marke saes. In he lieraure, Balcilar e al. (013) is he firs sudy o exend herding ess o a regime-swiching framework and heir resuls show ha herding ess based on he saic model can fail o idenify such behavior when herding is presen during a paricular marke sae only. For his purpose, we esimae a wo-sae Markov-wiching (M) model in he form CAD k, 0, 1, Rk,, Rk, e,, e h, ~ 0,1, h 1 N, 0, 1, 1,, 1, h e (3) where 1, follows a firs-order wo-sae M process. imilarly, h, sands for he sae-dependen condiional variance and is assumed o follow an independen swiching GARCH(1,1) process in order o avoid problems of recombining and analyical inracabiliy (Haas e. al., 004). In his specificaion,, measures he impac of unexpeced random shocks on volailiy in sae, whereas 1, and, ogeher measures he degree of sae-dependen volailiy persisence. If herd behavior is indeed presen during he high volailiy 9
sae only, =, hen one would expec α, o be negaive and significan and α,1 o be insignifican. On he oher hand, if commodiy financializaion is indeed a facor driving volailiy in commodiy prices, hen one migh argue ha shocks in he sock marke can also be a conribuing facor for herd behavior in he commodiy marke. Tha is, financial invesors correlaed rading aciviies moving funds across sock and commodiy markes may lead o a possible link beween large price movemens in he sock marke and herd behavior in he commodiy marke. Therefore, in order o es possible sock marke effecs on herd behavior in he commodiy marke, we modify Equaion (3) and esimae CAD k, 0, 1, R k,, R k, 3, R P, e, e, h,, 1 ~ N0,1 h h e (4), 0, 1, 1,, 1, where R P, is he reurn on he &P 500 index on day. A similar model is uilized by Chiang and Zheng (010) in order o examine he effec of he U.. marke on herd behavior in a number of global sock markes. In his model, observing a negaive and saisically significan esimae for α 3,s suggess ha large price movemens in he sock marke conribues o herd behavior in commodiy secor k during sae s. Following Kyle and Xiong (001), one can argue ha porfolio rebalancing of commodiy index funds can lead o correlaed rades in relaed markes and hus creae spillover effecs across differen commodiies. Furhermore, a number of sudies in he lieraure including Tyner (010), Alghalih (010), Du e al. (011), and ari e al. (01) documen causaliy and spillover effecs across commodiy secors, in paricular beween energy and agriculural secors. To his end, one migh argue ha he presence of herd behavior in a paricular commodiy secor can be associaed wih similar invesor behavior in anoher secor of he commodiy marke. Therefore, herding comovemens across differen commodiy secors can be observed eiher as a resul of 10
common risk facors driving commodiy reurns or hrough spillover effecs. Furhermore, following he commodiy financializaion hypohesis, one can also argue ha financial invesors rading aciviy, paricularly during periods of marke sress, may lead o correlaed rades across he differen commodiy secors and hus lead o an associaion of herd behavior across differen marke secors. For his purpose, we examine possible cross-herding effecs and esimae an augmened model of he form CAD k, 0, 1, R k,, R k, 3, CAD j, 4, R j, e, e, h,, 1 ~ N0,1 h h e (5), 0, 1, 1,, 1, where CAD j, and R j, are he cross-secional absolue deviaion and he reurn for he commodiy secor j on day, respecively. In his model, observing a negaive and saisically significan esimae for α 4,s suggess ha commodiy secor k ends o herd wih commodiy secor j during sae s. imilarly, observing a posiive and saisically significan esimae for α 3,s suggess he presence of co-varying risk associaed wih commodiy secors so ha a shock in secor j ends o be correlaed wih a shock in commodiy secor k. 4. Empirical resuls 4.1 Descripive saisics Panels A and B in Table 1 presen he descripive saisics for he average daily index reurns and he cross-secional absolue sandard deviaions of reurns (CAD) for each commodiy secor, respecively. All commodiy secors experienced posiive average reurns during he sample period wih he average reurn ranging beween a high of 0.038% for energy and low of 0.010% for livesock. On he oher hand, energy is he mos volaile secor followed by grains. Examining he higher momens, all commodiy reurns wih he excepion of grains are negaively skewed. Livesock secor has he smalles kurosis and volailiy. The highes level of reurn dispersion (Panel B) is observed in he energy secor, suggesing 11
higher marke variaions across energy fuures reurns, compared o oher commodiy secors. The high level of reurn dispersion observed may be due o unexpeced shocks observed in he energy secor, possibly driven by he uncerainy surrounding he energy marke due o a number of geopoliical issues and wars during much of he 000s. Livesock fuures, on he oher hand, exhibi he lowes level of reurn dispersion suggesing ha fuures conracs in his secor display greaer direcional similariy, hus leading o smaller reurn dispersion across fuures reurns in his commodiy secor. The low dispersion observed across livesock fuures reurns could be uilized in cross-hedging sraegies in his commodiy secor as low dispersion suggess greaer direcional similariy across he differen livesock conracs and hus greaer cross-correlaions wihin his commodiy secor. 4. Herding during low and high volailiy saes Table presens our findings for Equaions and 3 for he whole sample period. The esimaions are done using he common sample for he period beween Jan. 17, 1995 and Nov. 30, 01 wih 4,477 daily observaions for each commodiy secor. Consisen wih sandard asse pricing models, he models yield posiive esimaes for α 1,s (s=1,) for all commodiy secors, as he cross-secional variaion in asse sensiiviies leads o greaer reurn dispersion as each commodiy responds differenly o he marke reurn. In he case of herding ess, he saic model of Equaion rejecs herding for all commodiy secors. However, he regime-based specificaion yields suppor for herd behavior in grains during he high volailiy sae only (sae ) indicaed by a negaive and significan esimae for α,. As explained earlier, a non-linear and negaive relaion beween reurn dispersion and marke reurn suggess ha asse reurns display greaer direcional similariy during periods of large price movemens and, according o his mehodology, is consisen wih herd behavior. The finding of herd behavior during he high volailiy sae only is also consisen wih he basic raionale behind he esing mehodology ha invesors would be more likely o exhibi herding endencies during periods of marke sress. On 1
he oher hand, he findings rejec herding for he oher commodiy secors. In fac, he finding of no herding for energy and meals is consisen wih Pierdzioch e al. (010) and Pierdzioch e al. (013) who documen evidence of ani-herding among oil and meal price forecasers, respecively. Pierdzioch e al. (013) sugges ha ani-herding behavior reflecs a sraegy among forecasers driven by incenives o scaer forecass around a consensus forecas. In he volailiy equaion,, measures he sae-dependen impac of unexpeced random shocks on volailiy. We consisenly find ha he unexpeced random shocks have a larger impac on volailiy in he high volailiy sae (sae ). imilarly, we observe ha volailiy clusering is more pronounced in he high volailiy sae indicaed by greaer values for ). In he case of grains for insance, 1 is esimaed o be 0.17 and 0.665 for (,, 1 he low and high volailiy saes, respecively. Examining he esimaes across he commodiy secors, we find ha grains have he lowes volailiy clusering in he high volailiy sae wih a value of 0.665. Coupled wih he earlier finding of herd behavior in grains during he high volailiy sae only, he relaively low degree of volailiy clusering in his commodiy secor is consisen wih prior sudies suggesing ha herd behavior is a shor-lived phenomenon. Table 3 presens he esimaes for Equaion 4. The findings show ha large price movemens in he sock marke are generally associaed wih greaer reurn dispersions across commodiy reurns indicaed by posiive esimaes for α 3, (s=1,) in general. This suggess ha sock marke movemens have no significan herding effec on commodiies since herding would be evidenced by significanly lower dispersion across commodiy reurns during large marke movemens. This is indeed consisen wih he sandard asse pricing models suggesing ha asses would behave differenly during periods of large movemens as each asse would be differen in is sensiiviy o he marke reurn shock. On he oher hand, significan α 3 esimaes observed, paricularly in he case of meals, sugges ha correlaions among meal fuures reurns are significanly affeced by large price movemens in he sock marke as reurn dispersion and correlaion are 13
closely relaed. 4 The lack of a significan herding effec of he sock marke is consisen wih Adrangi and Charah (008), Buyuksahin and Harris (011) and een and Gjolberg (013) and provides suppor agains he financializaion of commodiies from a differen angle. Table 4 presens he findings for cross-marke herding effecs described in Equaion 5. In each panel, we focus on a arge commodiy secor and examine, in separae columns in he panel, he cross-herding effecs of he remaining hree commodiy secors described as he originaing secor where he cross-herding effec is assumed o originae from. The findings sugges ha energy and grains in general exhibi he greaes cross-marke sensiiviies. For example, in Panel A where he arge secor is energy, all oher commodiy secors are found o have significan cross-marke effecs wih negaive and significan α 4, esimaes, during he high volailiy sae only. imilarly, in he case of grains repored in Panel C, all commodiy secors are found o have negaive effecs alhough livesock is found o be insignifican. This suggess ha grains end o herd during he high volailiy sae around he energy and meals secors, suggesing an associaion beween herd behavior in grains and large price movemens in energy and meals secors. The finding of cross-commodiy marke herding effecs beween grains and energy fuures is consisen wih a number of prior sudies documening dynamic inerrelaionships beween energy and agriculural markes including Tyner (010), Du e al. (011), ari e al. (01), and Nazlioglu e al. (013). Our findings also show ha a cross-marke herding effec is presen from energy o grains. The findings also sugges ha posiions in energy and grains fuures will no provide significan diversificaion benefis in a porfolio as large price movemens in each commodiy secor would be associaed wih greaer direcional similariy across fuures reurns. The cross-marke dispersion effec esimaes α 3, (s=1,) in Equaion 5 do no sugges a consisen paern regarding he associaion of reurn dispersions across commodiy secors. Overall, he empirical resuls for he whole sample period sugges ha herd behavior is presen in grains 4 ee Demirer (013) for a more deailed analysis of he relaion beween reurn dispersion and correlaion. 14
during high volailiy sae wih significan cross-marke herding effecs across energy and grains. 4.3 The effec of he pos-004 period A number of sudies including Irwin and anders (01), Tang and Xiong (01), een and Gjolberg (013), among ohers, noe he dramaic increase in he open ineres and rading volume in he commodiy marke afer 003 due o he influx of financial invesors. Malkowski (011) noes a CFTC saff repor saing ha he oal value of various commodiy index relaed insrumens purchased by insiuional invesors increased from an esimaed $15b in 003 o a leas $00b in mid-008. The open ineres for seleced commodiies in Figure 1 clearly demonsraes he dramaic increase, paricularly afer 004. een and Gjolberg (013) documen evidence of increased co-movemens across commodiies afer 004, however hey conclude ha his resul is mainly driven by exreme price movemens during 008, suggesing no significan suppor for financializaion or conaminaion from financial invesor s aciviies. In order o check he robusness of our findings regarding he role of financializaion on herd behavior in commodiy markes, we modify Equaion 3 and esimae CAD k, 0, 1, R k,, R k, 1, D k, Rk, e, e, h,, 1 ~ N0,1 h h e (6), 0, 1, 1,, 1, where Dk, is a dummy variable ha akes on he value one saring wih January 1, 004. In his specificaion, observing a significan and negaive esimae for (α, +δ1,s) suggess he presence of herd behavior during he pos-004 period only. imilarly, Equaion 4 is modified as CAD e h k, 0, 1, R k,, R k, 1, D k, Rk, 3, R P,, D k, RP,, h,, 1 ~ N0,1 h e (7), 0, 1, 1,, 1, e, 15
in which he erm δ,s is used o es he possible herding effec of he sock marke during he pos-004 period. Table 5 presens he findings for Equaion 6. In general, he pos-004 period is found o have a negaive effec on reurn dispersions overall indicaed by negaive and highly significan δ1 esimaes. This suggess a significan srucural break in he relaionship beween he dispersion of commodiy reurns and marke reurn shocks afer 004 and is consisen wih he finding by een and Gjolberg (013) of increased co-movemens across commodiies during his period. However, examining he esimaes for (α, +δ1,s), we conclude ha herd behavior was no presen during his period, furher supporing our findings for he whole sample period. The findings for Equaion 7 presened in Table 6 lead o similar conclusions regarding he role of he sock marke during he pos-004 period, suggesing no significan herding effec of he sock marke during his period. Overall, our findings for he whole sample period as well as he pos-004 period do no yield suppor for he commodiy financializaion hypohesis. 5. Conclusions The main goal of his paper is o conribue o he debae on commodiy financializaion by exending ess of herd behavior o commodiy fuures markes. Uilizing daa from four commodiy secors including energy, grains, livesock and meals, we employ a reurn dispersion based esing mehodology exensively used in he lieraure o deec herd behavior. Our findings yield significan evidence of herd behavior in he marke for grains during he high volailiy sae only indicaed by significanly lower reurn dispersions across grains fuures reurns during periods of large price movemens. The finding of significanly lower reurn dispersions across grains fuures suggess ha cross-hedging sraegies using grains fuures may be uilized, paricularly during periods of high volailiy, as reurns in his commodiy secor would display greaer direcional similariy leading o lower dispersion. We also find ha large price movemens in energy and meals fuures significanly conribue o herding in grains, providing suppor for he 16
dynamic relaionship beween energy and agriculural commodiy reurns from a differen angle. I is possible ha volailiy ransmission across he energy and agriculural markes acs as a conribuing facor for herd behavior among invesors in he marke for grains. This finding also suggess ha combining asses from he energy, meals and agriculural commodiy secors in a porfolio will no provide significan diversificaion benefis as large price movemens in each commodiy secor would be associaed wih greaer direcional similariy across fuures reurns in he oher commodiy secors, hus eroding benefis from diversificaion. Consisen wih Adrangi and Charah (008), Buyuksahin and Harris (011) and een and Gjolberg (013), our findings for he whole sample period do no sugges ha herd behavior is presen in oher commodiy secors. The robusness checks for he pos-004 period during which he commodiy marke experienced a significan influx of financial invesors lead o similar resuls suggesing ha herd behavior is no presen in he oher commodiy secors. imilarly, our ess do no yield a significan sock marke effec on herd behavior in he commodiy marke boh during he whole sample period and he pos-004 period. The findings are consisen wih previous lieraure documening ani-herding among energy and meal marke forecasers. Overall, our findings do no provide any suppor for he commodiy financializaion hypohesis much debaed in he media as well as among policy makers. 17
References Adrangi, B. and Charah, A., 008. Do Commodiy Traders Herd? The Financial Review 43, 461-476. Alghalih, M., 010. The ineracion beween food prices and oil prices. Energy Economics 3, 150 15. Balcilar, M., Demirer, R., Hammoudeh,., 013. Invesor Herds and Regime-wiching: Evidence from Gulf Arab ock Markes. Journal of Inernaional Financial Markes, Insiuions & Money 3, 95-31. Bikhchandani,., Hirshleifer, D., Welch, I., 199. A heory of fads, fashion, cusom and culural change as informaional cascades. Journal of Poliical Economy 100, 99-106. Bikhchandani,., harma,., 001. Herd behavior in financial markes: A review. IMF aff Papers 47 (3), 79-310. Boyd, N., Buyuksahin, B., Haigh, M.., Harris, J. H., 009. The Prevalence, ources and Effecs of Herding. CFTC Working Paper. Buyuksahin, B. and Harris, J. H., 011. Do peculaors Drive Crude Oil Fuures Prices? The Energy Journal 3 (), 167-0. Chang, E. C., Cheng, J. W., Khorana, A., 000. An examinaion of herd behavior in equiy markes: An inernaional perspecive. Journal of Banking and Finance 4(10), 1651-1699. Chiang, T. C. and Zheng, D., 010. An empirical analysis of herd behavior in global sock markes. Journal of Banking and Finance 34 (8), 1911-191. Chrisie, W. G.., Huang, R. D., 1995. Following he pied piper: Do individual reurns herd around he marke? Financial Analys Journal, July-Augus, 31-37. Con, R., Bouchaud, J. P., 000. Herd Behavior and Aggregae Flucuaions in Financial Markes. Macroeconomic Dynamics, 170-196 Chunrong, A., Charah, A., ong, F., 006. On The Comovemen of Commodiy Prices. American Journal of Agriculural Economics, 88 (3), 574 588. Demirer, R., Kuan, A. M., 006. Does Herding Behavior Exis in Chinese ock Markes? Journal of Inernaional Financial Markes, Insiuions and Money 16, 13-14. Demirer, R., Kuan, A. M., Chen, C., 010. Do Invesors Herd in Emerging ock Markes? Evidence from he Taiwanese Marke. Journal of Economic Behavior & Organizaion 76, 83-95. 18
Demirer, R., 013. Can Advanced Markes Help Diversify Risks in Fronier ock Markes? Evidence from Gulf Arab ock Markes. Research in Inernaional Business and Finance 9, 77-98. Dennis, P., rickland, D., 00. Who blinks in volaile markes, individuals or insiuions? Journal of Finance 51, 111-135. Du, X., C.L., Yu, Hayes, D.J., 011. peculaion and volailiy spillover in he crude oil and agriculural commodiy markes: a Bayesian analysis. Energy Economics 33 (3), 497 503. Devenow, A., Welch, I., 1996. Raional herding in financial economics. European Economic Review 40, 603-615. Falkowski, M., 011. Financializaion of Commodiies. Conemporary Economics (5) 4, 4-17. Gabaix, X., Gopikerishman, P., Plerou, V., anley, H. E., 006. Insiuional invesors and sock marke volailiy. Quarerly Journal of Economics 11 (), 461-504. Gilber, C. L., 009. Commodiy peculaion and Commodiy Invesmen. Working paper No. 080, Deparmen of Economics, Universiy of Treno. Gleason, K. C., Lee, C., Mahur, I., 003. Herding behavior in European fuures markes. Finance Leers 1, 5-8. Haas, M., Minik,., Paolella, M.., 004. A New Approach o Markov-wiching GARCH Models. Journal of Financial Economerics, 493-530. Hamilon, J., 009. Causes and consequences of he oil shock of 007-008, Working paper, UC an Diego. Irwin,. H., anders, D. R., 01. Financializaion and rucural Change in Commodiy Fuures Markes. Journal of Agriculural and Applied Economics 44, 3(Augus), 371 396. Kilian, L., 009. No all oil price shocks are alike: Disenangling demand and supply shocks in he crude oil marke. American Economic Review 99, 1053-1069. Krugman, P., 008. Fuels on he hill. The New York Times (June 7). Luo, D., 003. Marke volailiy and muual funds cash flows. Working paper No. 03-1, Yale Inernaional Cener for Finance. Nazlioglu,., Erdem, C., oyas, U. 013. Volailiy spillover beween oil and agriculural commodiy markes. Energy Economics 36, 658-665. Philippas. N., Economou, F., Babalos, V., Kosakis, A., 013. Herding behavior in REITs: Novel ess and he role of financial crisis. Inernaional Review of Financial Analysis, forhcoming. Pierdzioch, C., Rülke, J. C., admann, G., 010. New evidence of ani-herding of oil-price 19
forecasers. Energy Economics 3, 1456 1459. Pierdzioch, C., Rülke, J. C., admann, G., 013. Forecasing meal prices: Do forecasers herd? Journal of Banking and Finance 37, 150 158. ari, M., Hammoudeh,., Chang C., McAleer, M., 01. Causaliy beween marke liquidiy and deph for energy and grains. Energy Economics 34, 1683 169. een, M., Gjolberg, O., 013. Are commodiy markes characerized by herd behaviour? Applied Economics 3, 79 90. Tan, L., Chiang, T. C., Mason, J. R., Nelling, E., 008. Herding behavior in Chinese sock markes: An examinaion of A and B shares. Pacific-Basin Finance Journal 16, 61 77. Tang, K., Xiong W., 01. Index invesmen and financializaion of commodiies, Financial Analyss Journal 68 (6), 54-74. Tyner, W.E., 010. The inegraion of energy and agriculural markes. Agriculural Economics 41, 193 01. Weiner, R. J., 006. Do Birds of a Feaher Flock Togeher? peculaor Herding in he World Oil Marke. Discussion paper No. 06-31. Resources for he Fuures. 0
Figure 1. Monhly open ineres for seleced commodiies. 1
Table 1. Descripive saisics for daily reurn dispersion and index reurns. All commodiies Energy Livesock Grains Meals Panel A: Index reurn Mean (%) 0.08 0.038 0.010 0.01 0.017 d. dev. (%) 1.453 1.998 0.91 1.507 1.180 Min. (%) -9.145-14.399-4.50-8.58-7.171 Max (%) 7.15 9.809 4.631 7.687 6.684 kewness -0.4-0.1-0.068 0.008-0.88 Kurosis 5.717 5.98 3.781 5.036 6.418 Panel B: Reurn dispersion (CAD) Mean (%) 1.360 1.076 0.793 0.853 1.039 d. dev. (%) 0.603 0.680 0.461 0.495 0.617 Min. (%) 0.374 0.087 0.035 0.017 0.091 Max (%) 7.33 5.763 4.697 4.760 5.507 kewness 1.947 1.935 1.433 1.667 1.93 Kurosis 10.88 9.187 6.901 8.143 9.147 Noe: The common sample covers he period Jan. 17, 1995 Nov. 30, 01 wih 4,477 observaions. CAD is he daily reurn dispersion wihin each commodiy secor as defined in Equaion (1) and he secor index is he &P GCI index for he corresponding commodiy secor.
Table. Herding behavior in he commodiy marke. All Commodiies Energy Livesock Grains Meals aic Regime-based aic Regime-based aic Regime-based aic Regime-based aic Regime-based Mean Equaion Mean Equaion Mean Equaion Mean Equaion Mean Equaion 0.935 0.848 0.879 0.75 0.667 0.505 0.651 0.574 0.803 0.678 (0.011)*** (0.014)*** (0.017)*** (0.03)*** (0.01)*** (0.014)*** (0.013)*** (0.014)*** (0.014)*** (0.014)*** 1.065 1.305 0.955 0.99 1.154 (0.0)*** (0.050)*** (0.0)*** (0.03)*** (0.039)*** 0.315 0.66 0.035 0.04 0.051 0.117 0.15 0.055 0.14 0.117 (0.015)*** (0.017)*** (0.017)*** (0.019)** (0.06)** (0.047)*** (0.016)** (0.03)** (0.0)*** (0.01)*** 0.383 0.093 0.011 0.190 0.37 (0.03)*** (0.038)*** (0.01) (0.033)*** (0.048)*** 0.037 0.011 0.017 0.007 0.06-0.009 0.003 0.015 0.04 0.015 (0.004)*** (0.004)*** (0.003)*** (0.003)** (0.011)*** (0.01) (0.004) (0.008)** (0.006)*** (0.006)** 0.038 0.016 0.106-0.0 0.031 (0.005)*** (0.006)*** (0.01)*** (0.006)*** (0.01)*** Volailiy Equaion Volailiy Equaion Volailiy Equaion Volailiy Equaion Volailiy Equaion β 0.005 0.063 0.013 0.17 0.00 0.059 0.011 0.055 0.010 0.085 (0.001)*** (0.005)*** (0.00)*** (0.008)*** (0.001)*** (0.004)*** (0.003)*** (0.076) (0.00)*** (0.005)*** β 0.003 0.01 0.000 0.110 0.008 (0.001)*** (0.014) (0.00) (0.046)*** (0.004)** β 0.99 0.000 0.89 0.000 0.960 0.008 0.875 0.17 0.90 0.000 (0.014)*** (0.043) (0.013)*** (0.070) (0.007)*** (0.057) (0.019)*** (1.094) (0.014)*** (0.09) β 0.958 0.91 0.97 0.54 0.97 (0.01)*** (0.046)*** (0.015)*** (0.158)*** (0.019)*** β 0.039 0.000 0.075 0.000 0.031 0.000 0.078 0.000 0.065 0.000 (0.007)*** (0.014) (0.008)*** (0.050) (0.005)*** (0.053) (0.010)*** (0.04) (0.009)*** (0.008) β 0.019 0.069 0.05 0.13 0.051 (0.007)*** (0.06)*** (0.007)*** (0.035)*** (0.011)*** LL -59.56-1931.04-3965.9-350.61-518.3-116.93-80.71-314.41-3439.71-967.84 Noe: Figures in parenheses are sandard errors and *, ** and *** indicae significance a 10%, 5% and 1%, respecively. LL sands for he likelihood value. 3
Table 3. The role of he sock marke on commodiy marke herding. All Commodiies Energy Livesock Grains Meals aic Regime-based aic Regime-based aic Regime-based aic Regime-based aic Regime-based Mean Equaion Mean Equaion Mean Equaion Mean Equaion Mean Equaion 0.930 0.894 0.871 0.747 0.667 0.504 0.646 0.574 0.784 1.138 (0.01)*** (0.07)*** (0.016)*** (0.018)*** (0.013)*** (0.014)*** (0.014)*** (0.014)*** (0.014)*** (0.039)*** 1.164 1.306 0.954 0.985 0.665 (0.034)*** (0.049)*** (0.06)*** (0.033)*** (0.014)*** 0.317 0.059 0.039 0.043 0.051 0.118 0.18 0.059 0.139 0.1 (0.015)*** (0.06) (0.015)*** (0.016)*** (0.06)** (0.09)*** (0.017)*** (0.03)*** (0.0)*** (0.047)*** 0.50 0.100 0.011 0.19 0.117 (0.059)*** (0.038)*** (0.03) (0.035)*** (0.00)*** 0.035 0.158 0.015 0.006 0.06-0.009 0.00 0.014 0.039 0.034 (0.004)*** (0.014)*** (0.003)*** (0.003)** (0.011)*** (0.013) (0.004) (0.007)** (0.006)*** (0.011)** 0.06 0.016 0.106-0.03 0.013 (0.009)*** (0.005)*** (0.015)*** (0.007)*** (0.006)** 0.004 0.016 0.007 0.005 0.000 0.000 0.003-0.00 0.017 0.019 (0.00)** (0.007)*** (0.00)*** (0.00)*** (0.001) (0.001) (0.00)** (0.00) (0.00)*** (0.003)*** 0.000 0.000 0.000 0.006 0.009 (0.003) (0.005) (0.003) (0.003)** (0.00)*** Volailiy Equaion Volailiy Equaion Volailiy Equaion Volailiy Equaion Volailiy Equaion β 0.005 0.06 0.013 0.008 0.00 0.035 0.011 0.056 0.010 0.013 (0.00)*** (0.098) (0.00)*** (0.003)*** (0.001)*** (0.03)* (0.003)*** (0.044) (0.00)*** (0.007)** β 0.005 0.007 0.000 0.103 0.083 (0.00)*** (0.013) (0.00) (0.040)*** (0.005)*** β 0.97 0.136 0.890 0.930 0.960 0.410 0.874 0.05 0.900 0.907 (0.015)*** (1.378) (0.013)*** (0.03)*** (0.007)*** (0.385) (0.01)*** (0.630) (0.015)*** (0.06)*** β 0.953 0.96 0.97 0.563 0.000 (0.014)*** (0.048)*** (0.015)*** (0.137)*** (0.017) β 0.040 0.01 0.077 0.004 0.031 0.000 0.079 0.000 0.064 0.057 (0.007)*** (0.014) (0.009)*** (0.00)*** (0.005)*** (0.006) (0.011)*** (0.010) (0.009)*** (0.013)*** β 0.07 0.064 0.05 0.11 0.000 (0.008)*** (0.030)** (0.007)*** (0.030)** (0.010) LL -57.11-010.3-3960.5-3513.6-518.3-116.78-818.43-311.16-3415.09-945.86 Noe: Figures in parenheses are sandard errors and *, ** and *** indicae significance a 10%, 5% and 1%, respecively. LL sands for he likelihood value. 4
Table 4. Herding effecs across commodiy secors. Panel A: Energy (Targe marke) Panel B: Livesock (Targe marke) Originaing Marke Originaing Marke Livesock Grains Meals Energy Grains Meals aic Regime aic Regime aic Regime aic Regime aic Regime aic Regime 0.859 0.769 0.87 0.74 0.859 0.754 0.650 0.97 0.645 0.938 0.64 0.475 (0.0) *** (0.00) *** (0.00) *** (0.03) *** (0.01) *** (0.018) *** (0.016) *** (0.036) *** (0.016) *** (0.05) *** (0.016) *** (0.041) *** 1.186 1.3 1.304 0.497 0.494 0.914 (0.060) *** (0.061) *** (0.13) *** (0.019) *** (0.017) *** (0.046) *** 0.034 0.04 0.034 0.038 0.035 0.04 0.05 0.015 0.049 0.005 0.053 0.119 (0.017) ** (0.014) *** (0.017) ** (0.016) *** (0.016) ** (0.015) *** (0.06) ** (0.041) (0.07) ** (0.019) (0.07) ** (0.10) 0.081 0.101 0.103 0.118 0.1 0.015 (0.034) *** (0.046) ** (0.037) *** (0.035) *** (0.03) *** (0.11) * 0.017 0.007 0.017 0.007 0.017 0.007 0.06 0.105 0.06 0.108 0.060-0.011 (0.003) *** (0.003) *** (0.003) *** (0.003) *** (0.003) *** (0.003) ** (0.011) *** (0.018) *** (0.01) *** (0.011) *** (0.01) *** (0.059) 0.018 0.016 0.016-0.009-0.013 0.104 (0.005) *** (0.006) *** (0.005) *** (0.017) (0.015) (0.041) *** 0.07-0.035 0.014 0.011 0.01-0.005 0.016 0.03 0.018 0.000 0.049 0.034 (0.017) * (0.016) ** (0.014) (0.030) (0.016) * (0.006) (0.009) * (0.01) (0.014) * (0.05) (0.011) *** (0.018) ** 0.155 0.005 0.014 0.006 0.003 0.045 (0.044) *** (0.09) (0.080) (0.009) (0.015) (0.030) * 0.001 0.006-0.00 0.000 0.001 0.005 (0.000) 0.000 0.004 0.007-0.005-0.00 (0.007) (0.006) (0.00) (0.00) (0.003) (0.004) * (0.001) (0.001) (0.001) *** (0.003) *** (0.00) *** (0.00) -0.008-0.011-0.008 0.000 0.003-0.005 (0.00) *** (0.005) *** (0.006) * (0.001) (0.001) ** (0.004) * LL -3964.76-351.05-3965.9-3515.43-3964.67-3516.18-516.83-115.55-51.46-110.97-507.51-109.05 5
Table 4 coninued aic Panel C: Grains (Targe marke) Originaing Marke Panel D: Meals (Targe marke) Originaing Marke Energy Livesock Meals Energy Livesock Grains Regimeswiching aic Regimeswiching aic Regimeswiching aic Regimeswiching aic Regimeswiching aic Regimeswiching 0.644 0.57 0.654 0.588 0.638 0.576 0.786 0.663 0.78 0.64 0.808 0.681 (0.017) *** (0.016) *** (0.015) *** (0.016) *** (0.017) *** (0.016) *** (0.018) *** (0.017) *** (0.019) *** (0.018) *** (0.017) *** (0.018) ** 0.99 0.954 1.030 1.166 1.078 1.177 (0.044) *** (0.04) *** (0.043) *** (0.045) *** (0.049) *** (0.060) *** 0.15 0.057 0.16 0.055 0.16 0.057 0.140 0.115 0.145 0.118 0.140 0.114 (0.016) *** (0.01) *** (0.017) *** (0.00) *** (0.016) *** (0.0) *** (0.0) *** (0.00) *** (0.0) *** (0.00) *** (0.0) *** (0.00) *** 0.189 0.191 0.185 0.34 0.0 0.17 (0.03) *** (0.033) *** (0.03) *** (0.050) ** (0.049) *** (0.050) *** 0.003 0.014 0.003 0.015 0.00 0.013 0.039 0.017 0.039 0.017 0.040 0.018 (0.004) (0.007) *** (0.004) (0.007) ** (0.004) (0.007) ** (0.006) *** (0.006) *** (0.006) *** (0.006) *** (0.006) *** (0.006) *** -0.01-0.03-0.019 0.034 0.038 0.036 (0.006) *** (0.006) *** (0.006) *** (0.01) *** (0.01) *** (0.01) *** 0.006-0.00-0.008-0.019 0.009-0.011 0.008 0.01 0.07 0.057-0.009 0.00 (0.009) (0.01) (0.010) (0.014) * (0.011) (0.013) (0.01) (0.009) * (0.016) *** (0.014) *** (0.010) (0.015) 0.016 0.055-0.05-0.013 0.086-0.031 (0.031) (0.034) * (0.09) (0.0) (0.043) ** (0.049) 0.000 0.001 0.003 0.001 0.003 0.008 0.00 0.001 0.017 0.008 0.001-0.00 (0.001) (0.001) * (0.005) (0.005) (0.00) (0.003) *** (0.001) ** (0.001) (0.006) *** (0.005) * (0.00) (0.00) *** -0.004-0.004-0.008 0.004 0.09 0.014 (0.00) ** (0.009) (0.004) ** (0.003)* (0.013) ** (0.004) *** LL -80.48-311.53-80.44-31.0-819.18-305.9-3436.79-964.7-340.58-949.47-3439.44-960.33 Noe: The able repors he esimaes for CAD k, 0, Rk Rk CAD 1,,,, 3, j, 4, R e. Figures in parenheses are sandard errors and *, ** and *** indicae significance a 10%, 5% and 1%, respecively. LL sands for he likelihood value. The volailiy equaion esimaes are no included for breviy and are available upon reques. j, 6
Table 5. The effec of he pos-004 period. All Commodiies Energy Livesock Grains Meals aic Regime-based aic Regime-based aic Regime-based aic Regime-based aic Regime-based Mean Equaion Mean Equaion Mean Equaion Mean Equaion Mean Equaion 0.940 0.840 0.887 0.755 0.653 0.90 0.643 0.573 0.809 0.688 (0.011)*** (0.01)*** (0.017) *** (0.019)*** (0.013)*** (0.08)*** (0.014)*** (0.013)*** (0.014)*** (0.014)*** 1.144 1.49 0.49 0.998 1.1 (0.05)*** (0.047)*** (0.013)*** (0.03)*** (0.048)*** 0.91 0.60 0.0 0.03 0.091 0.115 0.144 0.061 0.076 0.056 (0.014)*** (0.014)*** (0.016)* (0.017)* (0.07)*** (0.049)*** (0.018)*** (0.01)*** (0.03)*** (0.0)*** 0.308 0.18 0.118 0.18 0.159 (0.07)*** (0.036)*** (0.03)*** (0.03)*** (0.114)* 0.079 0.06 0.07 0.014 0.088 0.098-0.01 0.005 0.15 0.157 (0.004)*** (0.003)*** (0.003)*** (0.004)*** (0.01)*** (0.018)*** (0.006)** (0.007) (0.013)*** (0.03)*** 0.088 0.04 0.010-0.017 0.065 (0.007)*** (0.005)*** (0.010) (0.010)** (0.148) 1-0.056-0.051-0.011-0.006-0.081-0.083 0.013 0.015-0.10-0.16 (0.003)*** (0.003)*** (0.00)*** (0.00)** (0.008)*** (0.016)*** (0.004)*** (0.004)*** (0.010)*** (0.07)*** 1-0.051-0.019-0.037-0.006-0.03 (0.003)*** (0.004)*** (0.008)*** (0.008) (0.18) LL -089.38-181.76-3955.6-3507.7-469.49-097.01-815.4-307.5-3390.15-97.48 Noe: The able repors he esimaes for CAD k, 0, 1, Rk,, Rk, 1, Dk, Rk, e,. Figures in parenheses are sandard errors and *, ** and *** indicae significance a 10%, 5% and 1%, respecively. LL sands for he likelihood value. The volailiy equaion esimaes are no included for breviy and are available upon reques. 7
Table 6. The role of he sock marke during he pos-004 period. All Commodiies Energy Livesock Grains Meals aic Regime-based aic Regime-based aic Regime-based aic Regime-based aic Regime-based Mean Equaion Mean Equaion Mean Equaion Mean Equaion Mean Equaion 0.931 0.83 0.878 0.745 0.649 0.919 0.643 0.576 0.797 0.666 (0.011)*** (0.01)*** (0.016)*** (0.00)*** (0.013)*** (0.031)*** (0.014)*** (0.037)*** (0.014)*** (0.013)*** 1.140 1.38 0.490 0.990 1.189 (0.06)*** (0.046)*** (0.018)*** (0.087)*** (0.039)*** 0.96 0.67 0.03 0.08 0.090 0.117 0.144 0.061 0.069 0.078 (0.014)*** (0.014)*** (0.015)* (0.017)* (0.07)*** (0.056)*** (0.017)*** (0.064) (0.0)*** (0.00)*** 0.310 0.117 0.116 0.184 0.085 (0.09)*** (0.034)*** (0.04)*** (0.104)** (0.049)** 0.075 0.057 0.06 0.013 0.087 0.095-0.011 0.005 0.154 0.103 (0.004)*** (0.003)*** (0.003)*** (0.005)*** (0.01)*** (0.019)*** (0.005)** (0.07) (0.013)*** (0.010)*** 0.088 0.03 0.009-0.017 0.169 (0.008)*** (0.004)*** (0.03) (0.03) (0.09)*** 0.009 0.01 0.011-0.006 0.008 0.013-0.003-0.003 0.010 0.008 (0.00)*** (0.003)*** (0.004)*** (0.004)* (0.003)*** (0.010)* (0.00)* (0.004) (0.003)*** (0.003)*** 0.005 0.00 0.006 0.006 0.005 (0.005) (0.008)*** (0.003)** (0.041) (0.008) 1-0.055-0.050-0.01-0.008-0.078-0.078 0.011 0.015-0.105-0.080 (0.003)*** (0.003)*** (0.00)*** (0.003)*** (0.008)*** (0.033)*** (0.004)*** (0.011)* (0.010)*** (0.008)*** 1-0.054-0.016-0.035-0.006-0.11 (0.005)*** (0.004)*** (0.009)*** (0.008) (0.03)*** -0.005-0.009-0.004 0.013-0.008-0.013 0.009 0.00 0.014 0.00 (0.003)* (0.004)** (0.004) (0.005)*** (0.003)** (0.01) (0.003)*** (0.008) (0.004)*** (0.004) 0.00-0.01-0.006-0.001 0.01 (0.007) (0.009)** (0.003)** (0.043) (0.009)*** LL -08.13-1799.77-3947.33-3495.77-466.10-09.88-808.90-303.80-3363.61-905.45 Noe: The able repors he esimaes forcadk, 0, Rk Rk Dk Rk RP Dk RP e 1,,,, 1,,, 3,,,,,. Figures in parenheses are sandard errors and *, ** and *** indicae significance a 10%, 5% and 1%, respecively. LL sands for he likelihood value. The volailiy equaion esimaes are no included for breviy and are available upon reques. 8